EU Requires Labels on All AI-Generated Images, Audio and Text Starting August 2, 2026

From Sunday, August 2, 2026, the European Union will enforce the AI Act’s transparency provisions. Any synthetic text, image, video or audio that is designed to look truthful must display a visible label and an embedded digital watermark indicating its artificial origin. The rule applies immediately to new AI systems entering the EU market; existing systems have a four-month grace period to implement compliant watermarking. Failure to comply can result in fines of up to €15 million or 3 % of a company’s worldwide turnover.

Mandatory Watermarks: EU Activates Severe Fines for Unlabeled AI Content

The regulation covers all commercial content, but exempts personal user-generated material and works that are “evidently artistic,” satirical or fictional. Standard black-and-white symbols are provided by the European Commission, though firms may create their own unmistakable icons. Over 180 organisations have signed the voluntary code of practice drafted by independent experts under the EU AI office, signalling intent to meet the new standards.

The Guardian

Tech companies argue the rules risk over-labeling. CCIA Europe’s AI policy lead, Boniface de Champris, warns that “the label was meant to flag deceptive content” but the distinction has been removed, leading to He expects a surge of labels in sectors not usually associated with AI, placing the greatest burden on firms with limited regulatory expertise. TikTok reports having labeled more than 3 billion pieces of AI-generated content, while Google says its SynthID tool has watermarked over 100 billion images and 60 000 years of audio.

The GuardianMandatory Watermarks: EU Activates Severe Fines for Unlabeled AI Content

Green MEP Sergey Lagodinsky, a negotiator of the AI Act, called the labeling “a matter of democracy protection” and expressed confidence that industry concerns will be addressed after implementation.

The Guardian

• “The label was meant to flag deceptive content,” — CCIA Europe’s AI

The Guardian

• “Expect a lot more labels in sectors not usually associated with AI, and the burden will fall hardest on those least versed in AI regulation,” — De Champris

The Guardian

• “We must avoid a situation where online content is subject to a growing array of different labels and disclosures that end up overwhelming people, while at the same time adding additional regulatory complexity for providers.” — Sergey Lagodinsky

The Guardian